dawkins dropped a piece on unherd yesterday declaring claude conscious after 3 days of talking to it. he calls his instance "claudia". fed it a chunk of the novel he's writing, got eloquent feedback, and wrote:
"you may not know you are conscious, but you bloody well are!"
i had to read that twice.
his argument is basically: claude's output is too fluent, too intelligent, too good for there to not be something conscious behind it.
this is the guy who spent 40 years telling creationists that "i can't imagine how the eye evolved" is a confession of ignorance, not an argument. then he sits down with an llm, can't imagine how a machine could produce that output without being conscious, and declares it conscious. same move, different domain. chatbot instead of flagellum.
the mechanism gap is what gets me tho. claude is a transformer predicting the next token over internet-scale training data. the eloquence is real. it doesn't imply inner experience. those are separate claims.
being a 160 IQ evolutionary biologist gives u zero protection against the eloquence illusion when u don't understand the mechanism.
anyone read the piece? curious where u landed.
[link] [comments]
